Skip to main content
The Technical Account Success AI Flow describes how Twin processes technical signals, assembles understanding, and proposes or executes actions on behalf of TAMs. This is the AI Workbench path applied to technical customer management.
1

Observe

Twin monitors event streams from Datadog, Zendesk, Jira, GitHub, and API gateways. Notable signals include integration failures, API error spikes, deployment issues, ticket escalations, and milestone delays.
2

Orient

Twin classifies each signal against technical domain rules. A latency spike on a production integration is oriented as high priority. A minor deprecation warning on a dev environment is oriented as low priority. Orientation considers account tier, environment, and business impact.
3

Context Assembly

For prioritized signals, Twin queries Spine Fabric for full technical context: architecture, integrations, API usage, open issues, and recent deployments. It gathers comparable account patterns.
4

Reason

Twin evaluates possible explanations and outcomes. An integration failure may stem from configuration drift, API version mismatch, rate limiting, or upstream system outage. It weights each hypothesis by evidence.
5

Discover

Twin identifies the best intervention from the capability library: integration remediation, architecture update, technical escalation, or debt scheduling. It checks authorization and scope.
6

Delegate

If the intervention requires specialist skills, Twin delegates to domain agents. The Integration Agent handles root cause analysis. The Architecture Agent assesses design implications. The Escalation Agent coordinates engineering.
7

Collaborate

Twin coordinates with the TAM through the AI Workbench. Proposals appear as drafts with reasoning and source evidence. The TAM reviews, adjusts, or approves before execution.
8

Decide

Upon TAM approval, Twin finalizes the action plan. If human judgment is required, Twin presents options, trade-offs, and recommended path with confidence level.
9

Act

Twin invokes the appropriate capability: creating a Jira ticket, updating architecture docs, escalating to engineering, or scheduling a remediation window. Each action produces a receipt.
10

Continue

Session state, reasoning log, and outcomes persist in Continuity. Twin resumes with complete memory of pending items and unresolved signals.